Development of computer-aided diagnosis software for assessing tumor response to molecular targeted therapy by using a novel machine learning technique named CARTA

2016 Fiscal Year Final Research Report

• Project/Area Number: 25461864
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Radiation science
• Research Institution: National Cancer Center Japan
• Principal Investigator: Yamaguchi Masayuki 国立研究開発法人国立がん研究センター, 先端医療開発センター, ユニット長 (90450577)
• Co-Investigator(Renkei-kenkyūsha): KUTSUNA Natsumaro 東京大学, 新領域創成科学研究科, 准教授 (70578559) ; FUJII Hirofumi 国立がん研究センター, 先端医療開発センター 機能診断開発分野, 分野長 (80218982)
• Project Period (FY): 2013-04-01 – 2017-03-31
• Keywords: 画像診断学（含 放射線診断学） / がん分子標的治療 / 効果判定 / 橋渡し研究

Outline of Final Research Achievements

Currently, there is an immediate need for computer-aided diagnosis software that can assist radiologists accessing tumor response to cancer therapy, in particular molecular targeted therapy, because radiologists have to examine a quite large number of images in daily practice by using some complicated diagnostic criteria. We tested a novel machine learning technique named CARTA (Clustering-Aided Rapid Training Agent) for classifying MR images of various animal models, including pancreatic and colorectal cancer models. We acquired MR images of these animal models by using high-field (3.0 or 9.4 tesla) MR scanners. Our results showed that CARTA is a promising technique to classify MR images based on the similarity in image features. We contend that CARTA can produce software that can predict tumor response to various cancer therapies including molecular targeted therapy and radiation therapy based on MR image features.

Free Research Field

画像診断学